Skip to content

Fix: Triton inference spec hmac exposure#5654

Merged
nargokul merged 4 commits intoaws:masterfrom
pravali96:fix/triton-inference-spec-hmac-exposure
Apr 1, 2026
Merged

Fix: Triton inference spec hmac exposure#5654
nargokul merged 4 commits intoaws:masterfrom
pravali96:fix/triton-inference-spec-hmac-exposure

Conversation

@pravali96
Copy link
Copy Markdown
Collaborator

@pravali96 pravali96 commented Mar 19, 2026

Issue #, if available:

Description of changes:
fixes for HMAC key exposure and missing integrity check in the Triton inference handler and all model server prepare paths.

Issue

Fix

Switch from HMAC-SHA256 (requires a secret key) to plain SHA-256 (no key needed)

Changes

check_integrity.py

  • Removed generate_secret_key() — no longer needed
  • compute_hash() now uses hashlib.sha256() instead of hmac.new()
  • perform_integrity_check() no longer reads SAGEMAKER_SERVE_SECRET_KEY from environment

By submitting this pull request, I confirm that you can use, modify, copy, and redistribute this contribution, under the terms of your choice.

@pravali96 pravali96 force-pushed the fix/triton-inference-spec-hmac-exposure branch from 3b99611 to cb365fb Compare March 19, 2026 22:18
@pravali96 pravali96 force-pushed the fix/triton-inference-spec-hmac-exposure branch from 24f9bba to 30d0c21 Compare March 20, 2026 17:20
@pravali96 pravali96 changed the title Fix/triton inference spec hmac exposure Fix: Triton inference spec hmac exposure Mar 20, 2026
@pravali96 pravali96 force-pushed the fix/triton-inference-spec-hmac-exposure branch from 9e4d69c to 2db9ae8 Compare March 20, 2026 18:18
@pravali96 pravali96 force-pushed the fix/triton-inference-spec-hmac-exposure branch from 2db9ae8 to 8ecfe9d Compare March 20, 2026 19:34
@pravali96 pravali96 force-pushed the fix/triton-inference-spec-hmac-exposure branch from 01da135 to ad39a6c Compare March 23, 2026 16:02
Pravali Uppugunduri added 2 commits March 23, 2026 16:05
The ONNX export path in _prepare_for_triton() set self.secret_key to a
hardcoded value 'dummy secret key for onnx backend'. This key was then
passed as SAGEMAKER_SERVE_SECRET_KEY into container environment variables
and exposed in plaintext via DescribeModel/DescribeEndpointConfig APIs.

The ONNX path does not use pickle serialization — models are exported to
.onnx format and loaded natively by Triton's ONNX Runtime backend. There
is no serve.pkl, no metadata.json, and no integrity check to perform.
The secret key was dead code that also constituted a hardcoded credential
(CWE-798).

With this change, self.secret_key remains empty string (set by
_build_for_triton), and the existing cleanup in _build_for_transformers
removes empty SAGEMAKER_SERVE_SECRET_KEY from env_vars before CreateModel.

Addresses: P400136088 (Bug 2 - Hardcoded secret key)
Addresses P400136088 Bug 1 and V2146375387 (Triton path).

Three changes:

1. check_integrity.py: Switch from HMAC-SHA256 to plain SHA-256.
   - Remove generate_secret_key() — no longer needed
   - compute_hash() now uses hashlib.sha256() instead of hmac.new()
   - perform_integrity_check() no longer reads SAGEMAKER_SERVE_SECRET_KEY
     from environment

2. triton/model.py: Add integrity check in initialize() BEFORE
   cloudpickle deserialization. Previously the handler called
   cloudpickle.load() with no verification (acknowledged by a TODO
   comment). Now reads the file into a buffer, runs
   perform_integrity_check(), then deserializes with cloudpickle.loads().

3. triton/server.py: Remove SAGEMAKER_SERVE_SECRET_KEY from container
   environment variables in both local and SageMaker deployment modes.
   The key is no longer needed since integrity checking uses plain
   SHA-256.

4. model_builder_utils.py: Update _hmac_signing() to use plain SHA-256
   and stop generating/storing a secret key. Remove generate_secret_key
   import.

The integrity check still detects accidental corruption of model
artifacts in S3. The HMAC was providing a false sense of security since
the key was exposed via DescribeModel/DescribeEndpointConfig APIs.
@pravali96 pravali96 force-pushed the fix/triton-inference-spec-hmac-exposure branch from ad39a6c to ff838db Compare March 23, 2026 16:05
nargokul
nargokul previously approved these changes Mar 31, 2026
gwang111
gwang111 previously approved these changes Mar 31, 2026
@pravali96 pravali96 dismissed stale reviews from gwang111 and nargokul via ff838db March 31, 2026 22:55
@pravali96 pravali96 force-pushed the fix/triton-inference-spec-hmac-exposure branch from 08833d3 to ff838db Compare March 31, 2026 22:55
@pravali96 pravali96 force-pushed the fix/triton-inference-spec-hmac-exposure branch from ff838db to a341452 Compare March 31, 2026 22:55
Remove generate_secret_key import and usage from TorchServe, MMS,
TF Serving, and SMD prepare functions. Switch compute_hash calls
from HMAC-SHA256 to plain SHA-256 (no secret_key parameter).

This is required because generate_secret_key was removed from
check_integrity.py in the previous commit. Without this change,
all model server imports fail with ImportError.
@pravali96 pravali96 force-pushed the fix/triton-inference-spec-hmac-exposure branch from a341452 to c3744f2 Compare March 31, 2026 23:11
@nargokul nargokul merged commit 387213b into aws:master Apr 1, 2026
14 of 17 checks passed
@pravali96 pravali96 deleted the fix/triton-inference-spec-hmac-exposure branch April 1, 2026 17:00
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

3 participants